Some of us prefer DACs running on R2R topology and would like the equivalent in an analog-to-digital converter using eg successive approximation register (SAR) topology.
Would the 20-bit Linear LTC2378-20 work for audio sampling at 96k and 20-bits?
From the graphs in the datasheet at a sampling rate of 96kHz :
Would the 20-bit Linear LTC2378-20 work for audio sampling at 96k and 20-bits?
From the graphs in the datasheet at a sampling rate of 96kHz :
- Signal-to-(Noise + Distortion) Ratio ~ 102dBFS
- THD ~ 105dBFS
Since its a 1Msps part, you can run 10X oversampling to lower the noise - which for this part is about equivalent to a dithered 18bits. So perhaps close to 20bits will be possible.
It would probably work very good 🙂 It's THD is actually more like -125 db typical, the bad THD curve at higher frequencies is due to a crappy input buffer on the reference design...
The S/N ratio is also specified with a full bandwidth of 500 Khz, reducing it to the 20 Khz audio bandwidth should result in 113 db S/N.
Each time you decimate by 2 you get 1/2 bit increased resolution and gain 3 db improvement in S/N, so decimation by 8 gives you 9 db improvement in S/N, so it's is now at 122 db S/N....
And then there is the option of interleaving more than one chip, now 130 db S/N is in range....
Yes, as you might have noticed elsewhere I also don't like delta sigma converters and these newer very linear high speed SAR ADC chips can beat any delta sigma converter....
In fact I'm considering taking a couple of them and add a FPGA for the processing and filtering.... Could be useful for input ADC for a crossover or for an audio analyzer that can do bad things to the $28K APx555....
Hi all,
This would be very interesting to me as well. However, I wonder what such a converter's input stage really looks like? In both the LTC2378 & MAX11905 (SAR 1.6 Msps; 20 bit) datasheets there's a figure of a simplified input stage (fig 3 & 2 respectively), however, is the real world circuitry likely to sound superb? One thing is that the SAR algorithms may have advantages but the signal also has to get to the point where the SAR process "takes over" so to speak ...
Also, compared with a sigma delta converter it looks as if the input capacitance of an SAR converter is typically higher, potentially meaning that the spikes caused by capacitor charging is lower in a sigma-delta converter. My guess would also be that this may be more of an issue with the LTC2378 than the MAX11905 as the input resistance is higher in the MAX11905.
@soekris:
And then, just to return to my key question/consideration: Is the input stage/topology of an SAR converter likely to sound superb? One of you knows what such a topology looks like (maybe even for the 2378 or 11905)?
Cheers,
Jesper
This would be very interesting to me as well. However, I wonder what such a converter's input stage really looks like? In both the LTC2378 & MAX11905 (SAR 1.6 Msps; 20 bit) datasheets there's a figure of a simplified input stage (fig 3 & 2 respectively), however, is the real world circuitry likely to sound superb? One thing is that the SAR algorithms may have advantages but the signal also has to get to the point where the SAR process "takes over" so to speak ...
Also, compared with a sigma delta converter it looks as if the input capacitance of an SAR converter is typically higher, potentially meaning that the spikes caused by capacitor charging is lower in a sigma-delta converter. My guess would also be that this may be more of an issue with the LTC2378 than the MAX11905 as the input resistance is higher in the MAX11905.
@soekris:
Thanks for this piece of information soekris! I've been looking at the LTC2378 for some time but had actually omitted it due to this rise in HF distortion. But looking at the reference design (demo board: http://cds.linear.com/docs/en/demo-board-schematic/DC1925A2-SCH.pdf and the datasheet for the LT6203) your comment makes sense (assuming that part of the issue is that the LT6203's distortion is not at all low). Wonder why they made it like this ... ?
And then, just to return to my key question/consideration: Is the input stage/topology of an SAR converter likely to sound superb? One of you knows what such a topology looks like (maybe even for the 2378 or 11905)?
Cheers,
Jesper
A P.S.: Just came to think of that the aperture jitter for both the MAX11905 & LTC2378 looks to be quite high: 3 ps for the MAX11905 and 4 ps for the LTC2378. Higher than what I've normally seen discussed in the audio circles as "suitable" ...
Jesper
Jesper
I've got a parallella (linux on arm, fpga on chip and co-processor) and some of the header sockets (samtec) to build a daughter card for it with some of these ADCs next year 🙂
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
The input circuit of those SAR converters seems to be switched capacitors, therefore you need a large capacitor on the input pins and drive them from a low impedance driver.
The LTC2378 and MAX11905 is pretty much the same topology, main difference is that the MAX11905 is made in newer process technology, which unfortunate have the effect on lower input voltage, 3.3V instead of 5V, which everything else alike, lower the S/N 3 db. But then it's faster....
You could also look at 18 bits parts, then the AD7960 become very interesting....
How they sound like I have no clue, but with THD at -120db you can assume the errors are small. But I think they will sound better than any delta sigma ones with their typical 5th order modulating and noise shaping....
I'm considering doing small module boards with the different ADC part for testing, there just don't seems to be any info out there on the applications for audio use....
I'm not too worried about that for SAR ADC's as random jitter there will just generate noise.... And most audio ADC don't even specify jitter, so they're most likely not better.
There are also reports claiming you can't really hear below 2 nS jitter....
@soekris:
Hi Søren
Thanks for replying - interesting information & thoughts you share.
As it is I'm actually communicating with Maxim about the MAX11905. I have asked if they can provide a bit more information about the input stage of the 11905 so that I can get a grasp of how it - might - sound (given sensible layout and ancillaries). I've also invited Maxim to join in on this thread as I was thinking that they could have some interesting additional information. Time will show if they do so ...
But with your identification of the likely source of the HF distortion in the LTC2378 I am just now leaning more towards this AD converter - not least because it's much easier to work with timing wise than the MAX11905 (I'm not a programmer).
Hmmm... I reckon we have different perspectives here, Søren. As you may remember I sent you a PM a little while ago related to the Si514 that you - to my memory - use in your DAC. As it is one of the more significant eye-openers I've had in audio lately has been the effect of using a good/superb oscillator with my DAC (a DDDAC variant) - IMHO a superb clock establishes "the basis" for the level of detail, tonal nuance, spatiality, rhythm - many of the fundamental qualities of the music. That said, I respect that others/some research have a different opinion/other experiences.
Relative to the various ADCs eligible for audio I reckon you could be right in saying that the level of jitter may be similar with other ADCs including the delta-sigma types. It's interesting that such a - possibly - key parameter normally is not specified ...
@hochopeper:
Would indeed be interesting to me as well, hochopeper 😉
Cheers,
Jesper
Hi Søren
Thanks for replying - interesting information & thoughts you share.
As it is I'm actually communicating with Maxim about the MAX11905. I have asked if they can provide a bit more information about the input stage of the 11905 so that I can get a grasp of how it - might - sound (given sensible layout and ancillaries). I've also invited Maxim to join in on this thread as I was thinking that they could have some interesting additional information. Time will show if they do so ...
But with your identification of the likely source of the HF distortion in the LTC2378 I am just now leaning more towards this AD converter - not least because it's much easier to work with timing wise than the MAX11905 (I'm not a programmer).
Thanks also for the tip on this AD converter. But I'd like to stay with the 20 bit resolution (or higher if possible).
Hmmm... I reckon we have different perspectives here, Søren. As you may remember I sent you a PM a little while ago related to the Si514 that you - to my memory - use in your DAC. As it is one of the more significant eye-openers I've had in audio lately has been the effect of using a good/superb oscillator with my DAC (a DDDAC variant) - IMHO a superb clock establishes "the basis" for the level of detail, tonal nuance, spatiality, rhythm - many of the fundamental qualities of the music. That said, I respect that others/some research have a different opinion/other experiences.
Relative to the various ADCs eligible for audio I reckon you could be right in saying that the level of jitter may be similar with other ADCs including the delta-sigma types. It's interesting that such a - possibly - key parameter normally is not specified ...
@hochopeper:
Would indeed be interesting to me as well, hochopeper 😉
Cheers,
Jesper
Hi all,
Since the MAX11905 is interesting to me as an audio ADC I made contact with MAXIM's Applications support with some inquiries about its design. I have had some input from them that I find most valuable and - in case it is so to others as well - I post it here (with ok from Maxim):
"... because the front end is little more than what is shown there (in the datasheet, ed.): switches connected to capacitors, with some biasing and ESD protection. This is pretty standard on any modern SAR ADC without internal buffers. Very little, if any, sonic artifacts are likely to be generated here, from my experience - there are other circuit considerations that will have far more impact on the sound.
Second, recall that SAR converters were used at the dawn of digital audio. The SARs that used R2R DACs internally required sample/hold amplifiers - sometimes those were internal and others were external. The amplifier as well as the capacitors used were key to capturing a good sample, and many things could go wrong there (amplifier offset and drift, bias current, etc as well as cap leakage and voltco). These problems were almost all addressed by the shift to using CDACs instead of R2R in the SAR converter, because the CDAC acts as the sampling capacitor as well - and many things are done to trim the CDAC, minimize voltco, etc. It also removed an amplifier from the signal path.
Commenter "soekris" in the diyaudio thread makes an excellent point about the ADC driver circuitry being key. With the sampling capacitor now almost directly exposed, the signal must be buffered with circuitry that adequately charges the sampling cap as well as settles in the needed time. A poor driver design will do more to affect the conversion than anything that goes on inside an ADC with good linearity.
One reason the digital audio world moved to delta-sigma architectures was that they promised to remove the need for high-order active filtering prior to the ADC. Even with its 1MSPS sampling rate, the MAX11905 is going to need a 3rd to 5th order filter in front of it if you need a 20kHz bandwidth and expect any aliasing artifacts to be below 0.5LSB. The filters, and their realization, will color the sound more than the input circuitry of the ADC.
Lastly, the diyaudio discussion brought up concerns around the aperture jitter. While clock jitter is definitely something the system designer needs to ensure is low, the RMS noise voltage created by the 3ps aperture jitter of the MAX11905 is less than 1uV. The MAX11905 has ~6uV LSB; assuming a Gaussian noise distribution, the peak-to-peak noise resulting from aperture jitter is ~ 1LSB. You'll likely be filtering that out by digital filtering and decimation to achieve the processing gain discussed in the forum. So it's really not a major factor. The filters are the main concern.
The Maxim applications engineer concludes by saying that: " ... the MAX11905 was not designed with audio applications in mind nor have we done any listening tests with it .. "
Best regards,
Jesper
Since the MAX11905 is interesting to me as an audio ADC I made contact with MAXIM's Applications support with some inquiries about its design. I have had some input from them that I find most valuable and - in case it is so to others as well - I post it here (with ok from Maxim):
"... because the front end is little more than what is shown there (in the datasheet, ed.): switches connected to capacitors, with some biasing and ESD protection. This is pretty standard on any modern SAR ADC without internal buffers. Very little, if any, sonic artifacts are likely to be generated here, from my experience - there are other circuit considerations that will have far more impact on the sound.
Second, recall that SAR converters were used at the dawn of digital audio. The SARs that used R2R DACs internally required sample/hold amplifiers - sometimes those were internal and others were external. The amplifier as well as the capacitors used were key to capturing a good sample, and many things could go wrong there (amplifier offset and drift, bias current, etc as well as cap leakage and voltco). These problems were almost all addressed by the shift to using CDACs instead of R2R in the SAR converter, because the CDAC acts as the sampling capacitor as well - and many things are done to trim the CDAC, minimize voltco, etc. It also removed an amplifier from the signal path.
Commenter "soekris" in the diyaudio thread makes an excellent point about the ADC driver circuitry being key. With the sampling capacitor now almost directly exposed, the signal must be buffered with circuitry that adequately charges the sampling cap as well as settles in the needed time. A poor driver design will do more to affect the conversion than anything that goes on inside an ADC with good linearity.
One reason the digital audio world moved to delta-sigma architectures was that they promised to remove the need for high-order active filtering prior to the ADC. Even with its 1MSPS sampling rate, the MAX11905 is going to need a 3rd to 5th order filter in front of it if you need a 20kHz bandwidth and expect any aliasing artifacts to be below 0.5LSB. The filters, and their realization, will color the sound more than the input circuitry of the ADC.
Lastly, the diyaudio discussion brought up concerns around the aperture jitter. While clock jitter is definitely something the system designer needs to ensure is low, the RMS noise voltage created by the 3ps aperture jitter of the MAX11905 is less than 1uV. The MAX11905 has ~6uV LSB; assuming a Gaussian noise distribution, the peak-to-peak noise resulting from aperture jitter is ~ 1LSB. You'll likely be filtering that out by digital filtering and decimation to achieve the processing gain discussed in the forum. So it's really not a major factor. The filters are the main concern.
The Maxim applications engineer concludes by saying that: " ... the MAX11905 was not designed with audio applications in mind nor have we done any listening tests with it .. "
Best regards,
Jesper
Still considering products, but note that Linear Technology now have the LTC2380-24, a 24 bit, 1.5/2 Msps, -117 db THD part....
They also have the new and very impressive LTC2387, a 18 bit, 15 Msps, -117 db THD part....
Both are of course SAR ADC, no delta sigma....
Not quote right, both are the ADC equivalent to a Ladder DAC, the SAR ADC is just a serial version of the Flash ADC, the will sound the same, no delta sigma modulation....
Perhaps in the peculiar parallel universe where audiophiles dwell that may well be true but in electronic terms they are two different approaches.
- Status
- Not open for further replies.